Searching similar trajectories by locations

Abstract

Techniques for providing a trajectory route to multiple geographical locations of interest are described. This disclosure describes receiving global position system (GPS) logs associated with respective individual devices, each of the GPS logs including trajectories connecting a set of geographical locations previously visited by an individual of a respective individual device. A trajectory route service receives a request for a trajectory connecting a set of geographical locations of interest specified by a user. The trajectory route service calculates a proximal similarity between (1) the set of geographical locations of interest specified by the user, and (2) respective sets of geographical locations from the GPS logs. The trajectory route service constructs the requested trajectory with use of at least one of the trajectories from the GPS logs determined at least in part according to the calculated proximal similarities.

Description

BACKGROUND

A wide range of map-based services is being offered to users through web browsers, search engines, and as applications. Users may access the map-based services for street maps and a route planner for traveling by foot, vehicle, or public transport.

Requests for map-based services have become a common activity in people's daily lives. Many users access these services through a search engine on a computing device or on a personal navigation device. Users often request map-based services prior to driving to an appointment or an event at an unfamiliar location. However, conventional map-based services often determine routes solely with reference to a starting location to a destination location.

Other services, meanwhile, attempt to find user-requested routes using conventional trajectory searches. However, these trajectory map-based services often output planned routes that are based on a shape, a shape skeleton, a comparison, or other criteria. As such, these routes do not necessarily end precisely at a desired geographical location.

SUMMARY

This disclosure describes providing a trajectory route based on user input for multiple geographical locations. A trajectory route service receives global position system (GPS) logs (or other location-based logs) associated with respective devices, each of the GPS logs including trajectories connecting a set of geographical locations previously visited by an individual using a respective device. Next, a user requests a trajectory connecting a set of geographical locations of interest specified by the user. The trajectory route service calculates a proximal similarity between (1) the set of geographical locations of interest specified by the user, and (2) respective sets of geographical locations from the GPS logs. Based at least in part on the calculated proximal similarity, the trajectory route service constructs the requested trajectory with use of at least one of the trajectories from the GPS logs.

In another implementation, a trajectory route service receives a user input specifying multiple geographical locations of interest for planning a travel route. The trajectory route service accesses a trajectory route map constructed from GPS logs associated with respective individual devices, each of the GPS logs include trajectories that connect a set of geographical locations previously visited by an individual of a respective individual device. The trajectory route service computes an initial route by identifying trajectories from the GPS logs being closest in distance to each of the geographical locations of interest. The trajectory route service then refines the initial route by pruning and removing unqualified trajectories. The trajectory route service presents a route with a trajectory from the GPS logs that sequentially connects each of the multiple geographical locations of interest.

In yet another implementation, the trajectory route service receives a request for directions to multiple geographical locations and an order of travel to the multiple geographical locations. The trajectory route service presents a travel route in the order of travel, as specified by the user, to each of one or more geographical locations of interest.

This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

BRIEF DESCRIPTION OF THE DRAWINGS

The Detailed Description is set forth with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different figures indicates similar or identical items.

FIG. 1 illustrates an architecture to support an illustrative environment for providing a trajectory route to a user.

FIG. 2 is a flowchart showing illustrative phases for providing the trajectory route. The phases include preprocessing data, identifying trajectories that are similar, determining a trajectory connecting each of the geographical locations of interest, and constructing the trajectory route for use by the architecture of FIG. 1.

FIG. 3 illustrates an illustrative process of preprocessing GPS log data.

FIG. 4 is a flowchart showing an illustrative process of identifying trajectories from the GPS logs that are similar to the geographical locations of interest.

FIG. 5 illustrates examples of diagrams of matching points of a trajectory from the GPS logs to each of one or more geographical locations.

FIG. 6 is a flowchart showing an illustrative process of calculating a proximal similarity between (1) the set of geographical locations of interest specified by the user, and (2) respective sets of geographical locations from the GPS logs.

FIG. 7 is a flowchart showing an illustrative process of refining and pruning unqualified trajectories in constructing a requested trajectory or a travel route.

FIG. 8 illustrates an illustrative process of constructing a travel route in response to receiving user input.

FIGS. 9-10 illustrate example trajectory routes.

FIG. 11 is a block diagram showing an illustrative server usable with the architecture of FIG. 1.

DETAILED DESCRIPTION

Overview

This disclosure describes techniques for providing a travel route between a set of locations specified by a user. For instance, the user may desire to receive directions to a set of locations (e.g., locations A, B, and C), possibly without specifying a particular order of travel for the directions to these locations. After these desired locations are provided to a trajectory route service, the trajectory route service identifies an ideal route for travel to the multiple locations based in part on a travel sequence to the locations previously visited by individuals. To be able to provide the ideal route, the trajectory route service calculates a distance between the locations previously visited by individuals and the set of locations specified by the user. Thus, the trajectory route service presents the ideal route to the locations specified by the user based on the travel sequence from historical data.

In another implementation, the trajectory route service may further include an order of travel as specified by the user to one or more of multiple locations (e.g. locations A, B, and C). The user may desire directions starting at location A, travelling from location A to location C, and then travelling from location C to location B. For example, the user may want to view material at a new fabric store, attend a recital at a school, and meet friends for dinner at a restaurant that the user has not previously dined at. The trajectory route marks a travel route in the order as specified by the user, such as starting at location A, travelling from locations A to C, and travelling from locations C to B. Thus, the trajectory route service adjusts the travel sequence to each one of the locations of interest to satisfy the order of travel. The terms geographical locations of interest specified by the user and geographical locations of interest may be used interchangeably to describe the locations specified by the user.

As described herein, a trajectory route service constructs a trajectory or a travel route based on a relationship between global positioning system (GPS) trajectories and geographical locations of interest. The trajectory route service receives GPS logs associated with respective devices, each of the GPS logs including trajectories connecting a set of geographical locations previously visited by an individual of a respective device. The trajectory route service receives a request for a trajectory connecting the multiple geographical locations specified by a user. The trajectory route service calculates a proximal similarity between the set of geographical locations of interest specified by the user, and respective sets of geographical locations from the GPS logs.

While aspects of described techniques can be implemented in any number of different computing systems, environments, and/or configurations, implementations are described in the context of the following illustrative computing environment.

Illustrative Environment

FIG. 1 illustrates an illustrative architectural environment 100 in which a trajectory or a travel route may be recommended for output on a computing device. The environment 100 includes an illustrative computing device 102, which is illustrated as a personal digital assistant (PDA). The computing device 102 is configured to connect via one or more network(s) 104 to access a trajectory route service 106 on behalf of a user 108. The computing device 102 may take a variety of forms, including, but not limited to, a portable handheld computing device (e.g., a personal digital assistant, a smart phone, a cellular phone), a personal navigation device, a laptop computer, a desktop computer, a portable media player, or any other device capable of connecting to one or more network(s) 104 to access the trajectory route service 106 for the user 108.

The network(s) 104 represents any type of communications network(s), including wire-based networks (e.g., public switched telephone, cable, and data networks) and wireless networks (e.g., cellular, satellite, WiFi, and Bluetooth).

The trajectory route service 106 represents an application service that may be operated as part of any number of online service providers, such as a search engine, a map service, a social networking site, or the like. Also, the trajectory route service 106 may include additional modules or work in conjunction with other modules to perform the operations discussed below. In an implementation, the trajectory route service 106 may be implemented at least in part by a trajectory route application 110 executed by trajectory route servers 112, or by a trajectory route application stored in memory of the computing device 102. Updates may be sent for the trajectory route application stored in memory of the computing device 102.

The trajectory route service 106 may be hosted on one or more trajectory route servers, such as server 112(1), 112(2), . . . , 112(S), accessible via the network(s) 104. The trajectory route servers 112(1)-(S) may be configured as plural independent servers, or as a collection of servers that are configured to perform larger scale functions accessible by the network(s) 104. The trajectory route servers 112 may be administered or hosted by a network service provider that provides the trajectory route service 106 to and from the computing device 102.

In the illustration, the computing device 102 includes a trajectory route user interface (UI) 114 that is presented on a display of the computing device 102. The trajectory route service 106, in operation with the trajectory route application 110, presents the UI 114 to receive user input and to present the trajectory or the travel route to the user 108. Thus, the UI 114 facilitates access to the trajectory route service 106 that provides the trajectory or the travel routes.

In an implementation, the UI 114 is a browser-based UI that presents a page received from the trajectory route service 106. The user 108 employs the trajectory route UI 114 when accessing the trajectory route service 106 to find a map for a particular region. In an implementation, the UI 114 may allow the user 108 to select one or more geographical locations of interest on the particular region in the map by clicking on these locations. In response, the trajectory service may determine a best trajectory amongst these locations, as discussed in detail below.

In another implementation, the trajectory route service 106, in operation with the trajectory route application 110, presents the UI 114 to receive textual or aural input from the user 108. For instance, the user 108 may type one or more geographical locations of interest and, in response, the trajectory service may determine a best trajectory amongst these locations. In the illustrated example, the user 108 may input multiple geographical locations of interest without any travel order in which the user would like to visit these locations. For example, the UI 114 illustrates a location of “23 Peach St.,” another location identified by its landmark name such as “The Capitol,” and another location “9 Main St.” The trajectory route service 106 provides the trajectory or the travel route based on the trajectories from the GPS logs that are closest to each one of the geographical locations of interest, identified by a street name, a landmark name, or a specific point location that is of use or of interest to the user 108. The geographical location of interest or the specific point location may include but is not limited to a type of a location, such as a beach, a highway, a park, a camp site, an arena, a stadium, a name of an attraction, a name of a landmark, a name of a building, a name of an education facility, a street address, and the like. Furthermore, a number of geographical locations of interest that may be requested for one travel route may be ten or less.

In yet another implementation, the UI 114 may receive a request from the user 108 for a trajectory that is based on a specific travel sequence for the multiple geographical locations of interest. For instance, the user 108 may specify that she would like to visit “23 Peach St.” first, before visiting “The Capitol” second, and then “9 Main St.” In this instance, each one of the locations may have a number identifying a particular order of travel for the locations. That is, a first location may be indicated as “location 1 (L1)”, a second as “location 2 (L2)”, and so forth to show a desired travel sequence. In some instances, the order is specified explicitly by the user or by another user (e.g., a travel agent). In other instances, meanwhile, the order is determined based on other factors, such as a bus route, a travel agency's itinerary, traffic flow patterns of one way streets, traffic patterns, and the like.

In the illustration, the user 108 accesses the trajectory route service 106 via the network 104 using their computing device 102. The trajectory route service 106 presents the UI 114 to receive user input for geographical locations of interest and/or to provide the trajectory or the travel route for the multiple geographical locations of interest. In an implementation, the user 108 accesses a trajectory map for a particular region. Upon activating the particular region on the map, the user 108 may select the geographical locations of interest in the particular region, and the trajectory route service 106 provides a marked track for the trajectory route. In other implementations, the trajectory route may be used to plan daily routes, to plan for vacations, to analyze traffic flow patterns, to survey popular routes through attractions, to locate trajectories that are nearest to desired stationary places, and the like.

FIG. 2 is a flowchart showing an illustrative overview process 200 of high level functions performed by the trajectory route service 106. The process 200 may be divided into four phases, an initial phase to preprocess GPS logs 202 (or other location-based logs), a second phase to identify trajectories from GPS logs that are similar to the geographical locations of interest 204, a third phase to determine a best trajectory 206, and a fourth phase to construct a route to connect the multiple geographical locations of interest 208. All of the phases may be used in the environment of FIG. 1, may be performed separately or in combination, and without any particular order.

The first phase is to preprocess raw GPS logs to represent geographical locations 202. The trajectory route service 106 receives the GPS logs associated with respective individual devices. Each of the GPS logs includes trajectories connecting a set of geographical locations previously visited by an individual of a respective individual device.

The second phase is to identify trajectories from the GPS logs that are similar to the geographical locations of interest 204. The trajectory route service 106 calculates a proximal similarity between the set of geographical locations of interest and respective sets of geographical locations from the GPS logs.

The third phase is to determine the best trajectory connecting to each of the multiple geographical locations of interest 206. The trajectory route service 106 provides the best trajectory that connects the geographical locations of interest.

The fourth phase is to construct a route of a trajectory that allows requesting user to travel in order specified 208. The trajectory route service 106 provides the route in a travel sequence or allows a requesting user to travel to the multiple geographical locations of interest in an order specified by the user. Details of the phases are discussed in FIGS. 3-11 below.

Exemplary Processes

FIGS. 3, 4, 6, 7, and 8 are flowcharts showing illustrative processes for the phases mentioned above. The processes are illustrated as a collection of blocks in logical flowcharts, which represent a sequence of operations that can be implemented in hardware, software, or a combination. For discussion purposes, the processes are described with reference to the computing environment 100 shown in FIG. 1. However, the processes may be performed using different environments and devices. Moreover, the environments and devices described herein may be used to perform different processes.

For ease of understanding, the methods are delineated as separate steps represented as independent blocks in the figures. However, these separately delineated steps should not be construed as necessarily order dependent in their performance. The order in which the process is described is not intended to be construed as a limitation, and any number of the described process blocks maybe be combined in any order to implement the method, or an alternate method. Moreover, it is also possible for one or more of the provided steps to be omitted.

Preprocess GPS Log Data

FIG. 3 illustrates illustrative process for performing the phase 202 of preprocessing GPS log data. At 302, the trajectory route service 106 obtains, collects, or receives raw GPS log data (or other location based log) of geographical locations of individual users. The logs 302 may be obtained from GPS devices, tracking units, mobile phones, or any other device, as long as these devices are located in close proximity to each of the individuals. The GPS devices may be set to automatically track the position of the GPS device at regular intervals. Each of the individual users is asked prior to tracking or to sharing their movements and locations if their GPS data may be included as a part of a GPS data collection. The options are that the individual user may choose to opt-in to participate or to opt-out to not participate in the tracking or sharing of their GPS data. As such, the techniques track the GPS data after receiving user consent.

In another implementation, the trajectory route service 106 may obtain GPS logs from GPS-log driven applications, social networks, or services on the web. Each individual user may be equipped with a GPS device for tracking data. The device may include a GPS navigation device, a GPS phone, or any other type of GPS sensor that collects GPS log data at a high sampling rate, such as every two to eight seconds per point. The GPS data may be uploaded to the web by the users to show their positions and to share their GPS locations by agreeing to opt-in to participate in the data collection.

The GPS log 302 is generally a collection of a series of points represented points containing a latitude (Lat), a longitude (Lngt) and a time (T).

The trajectory route service 106 sequentially connects the points into a GPS trajectory 304. The trajectory 304 may be represented by:R=(p₁,p₂. . . ,p_n)Where n is a number of points in the trajectory, n=8.Identify Points on Trajectory to Represent Geographical Locations

FIG. 4 illustrates an illustrative process of the phase 204 of identifying points on a trajectory from the GPS logs to represent the geographical locations of interest. The trajectory route service 106 receives a set of trajectories from GPS logs 402, as described in FIG. 3.

Next, the trajectory route service 106 receives a request for a set of geographical locations of interest 404 specified by the user 108 through user input on the map or by text. The set of geographical locations of interest specified by the user 108 may be represented as:Q={q₁,q₂, . . . ,q_m}

where m is a number of locations. The geographical locations of interest Q may be assigned with a travel sequence, if specified by the user 108. If there is a travel sequence, Q is treated as a sequence of locations from q₁ to q_m.

The trajectory route service 106 searches for trajectories from the GPS database that are similar to the geographical locations of interest 406. In order to identify how well a trajectory from the GPS logs connects the geographical locations of interest, a distance (e.g., spatial) is measured with a value or distance amount and a similarity function or a proximal similarity are calculated. For example, at least one trajectory from the GPS logs is determined based on the at least one trajectory connecting a set of geographical locations from the GPS logs having a highest calculated proximal similarity to the set of geographical locations of interest specified by the user

The trajectory route service 106 calculates a spatial distance from the trajectory from the GPS logs to each geographical location of interest 408. The trajectory route service 106 calculates the spatial distance by using the following equation:

${\mathrm{Dist}}_q\left(q_i,R\right)=\underset{p_j\in R}{\min }\left\{{\mathrm{Dist}}_e\left(q_i,p_j\right)\right\}$ where R represents the trajectory, R={p₁, p₂, . . . , p_l}. On the right side of the equation, Dist_e(q_i, p_j) represents an Euclidean distance between a location of interest, q_i and a trajectory point, p_i. The Euclidean distance, Dist_e(q_i, p_i) is a measured amount of distance from q_i to any point p_i on R. If a small, short, or a closest distance has been identified, the <q_i, p_i> is referred to as a matched pair where p_j is a nearest point on R to q_i. However, p_j may be matched with multiple geographical locations.

The trajectory route service 106 evaluates the similarity function or the proximal similarity between the set of geographical locations of interest and the respective sets of geographical locations from the GPS logs, based on the trajectory from the GPS logs 410. The similarity is evaluated by using the following equation:Sim(Q,R)=Σ_i=1^me^{−Distq(qi,R)}.

The exponential function e^{−Distq(qi,R)} is used to measure a contribution of each matched pair to Sim(Q, R). The contribution from each matched pair may be determined by assigning weights to each pair. A larger weight may be assigned to a matched pair that is closer in distance while a smaller weight or less weight may be assigned to a matched pair that is further apart. The assigned weight results in an exponential decreasing of contribution as Dist_q(q_i,R) increases. Based on the weight assigned, the trajectory from the GPS logs that is within a threshold distance to each of the geographical locations of interest is considered to be “similar.”

In an implementation, the trajectory route service 106 may determine whether a distance between the geographical location of interest and the point in the trajectory from the GPS logs is greater than or less than a predetermined threshold. In an event that the distance is greater than the predetermined threshold, the trajectory route service 106 will refrain from including the trajectory in a candidate set of GPS trajectories. In an event that the distance is less than the predetermined threshold, the trajectory route service 106 may include the trajectory in the candidate set of GPS trajectories.

FIG. 5 illustrates an illustrative process 500 of identifying trajectories by calculating the closest distance of the trajectory points to the geographical locations of interest. The points on trajectory R are shown as R=p₂, p₃, p₄, p₅, p₆, p₇, p_n). For example, the geographical locations of interest are identified as q1, q2, and q3, represented by oval shapes in a hatchet pattern in 500. The geographical locations of interest identified as q1, q2, and q3 are matched to the closest trajectory points, p₆, p₄, and p₇, respectively as shown at 500. The dashed ellipses illustrate the matched pairs (e.g., matched pairs are points of the trajectory from the GPS logs closest to the geographical locations of interest). For example, the distance values for the matched pairs are: Dist_e(q₁, p₆)=1.5, Dist_e(q₂, p₄)=0.1, and Dist_e(q₃, p₇)=0.1. Thus, the similarity function may be represented as: Sim(Q, R)=e^{−Diste(q1, p6)}+e^{−Diste(q2, p4)}+e^{−Diste(q3, p7)}=e^−1.5+e^−0.1+e^−0.1.

The distance measurement for Dist_e(q₁, p₆)=1.5 is shown at 502.

Proceeding to 504 is an illustrative travel sequence specified by the user 108 for the geographical locations of interest. The matched points on the trajectory from the GPS logs may help satisfy the order that is specified by the user. However, the trajectory route service 106 may adjust the matched points, based on a travel sequence to the geographical locations in the order specified by the user 108. For example, the matching no longer occurs since a matched point p_j for the geographical location of interest q_i may not be the nearest point to q_i any longer. For example, the user 108 may specify the travel sequence of the locations, represented as q₁→q₂→q₃. However, the actual visiting order of the matched points on R is from p₄→p₆→p₇, assuming that the trajectory, R travels from left to right. The travel sequence is no longer from p₆→p₄→p₇. These matched pairs no longer conform to the user specified order, causing the trajectory route service 106 to adjust the matching of trajectory points to satisfy the order of travel requested by the user 108.

Shown at 504, q₁ is re-matched with p₃ and the new travel sequence is from p₃→p₄→p₇, which satisfies the user-specified order. The goal is to maximize a sum of the contribution of each matched pair, based on the weights, while still keeping the order of visits. The sum of the contribution of the pairs, <q₁, p₃>, <q₂, p₄>, and <q₃, p₇> is maximized among all of the possible combinations that satisfy the order of travel.

For the order specified by the user 108, an equation to calculate the similarity function with order Sim_o(Q, R) for the geographical locations of interest is:

${\mathrm{Sim}}_o\left(Q,R\right)=\max \{\begin{array}{c}{e}^{-{\mathrm{Dist}}_e\left(\mathrm{Head}\left(Q\right),\mathrm{Head}\left(R\right)\right)}+\\ {\mathrm{Sim}}_o\left(\mathrm{Rest}\left(Q\right),R\right)\\ {\mathrm{Sim}}_o\left(Q,\mathrm{Rest}\left(R\right)\right)\end{array}$ where Head(*) is a first point of *, where Head(Q)=q₁ and Rest(*) indicates that a rest part of * after removing the first point, e.g., Rest(Q)={q₂, q₃, . . . , q_m}. The equation for Sim_o(Q, R) defines maximal solutions for subproblems: Sim_o(Rest(Q), R) and Sim_o(Q, Rest (R)). Therefore, once Head(Q) and Head(R) match, e^{−Diste(Head(Q),Head(R))} may be summed up to the similarity function and shift to the matching of the rest of Q by using Sim_o(Rest (Q), R). Head(R) is retained for a next round of comparison as a trajectory point may be matched with more than one geographical locations of interest.

Dynamic programming is used to solve the similarity and to keep the matched trajectory points in a same order as the geographical locations of interest.

The equation to evaluate the similarity function is:Similarity(Q,R_i)_Ri∈T′≧Similarity(Q,R_j)_Rj∈T-T′where Similarity (Q, R_i)=Sim(Q, R_i) if no order is specified. If there is an order-specified, a subscript of o is used, Similarity (Q, R_i)=Sim_o(Q,R_i).

The trajectory route service 106 may search for trajectories from the GPS logs by retrieving trajectory points from the GPS logs that are within a threshold distance to each of the multiple geographical locations of interest, a trajectory point represents a geographical location previously visited by the user of a respective user device. The trajectory route service 106 identifies the retrieved trajectory points that are within an intersection of the multiple geographical locations of interest. Furthermore, the trajectory route service 106 may determine that the trajectory points that are within the intersection as being closest in distance to the multiple geographical locations of interest.

Identifying “Best” Trajectory

FIG. 6 illustrates an illustrative process 206 of determining the best trajectory to connect the geographical locations of interest. The trajectory route service 106 may use a spatial index that utilizes a rectangle-tree (R-tree) to search for trajectories from the GPS logs, followed by an in-depth study on further adaption and optimization. The R-tree includes tree data structures for indexing multi-dimensional information, such as the coordinates of the geographical locations. The R-tree splits space with hierarchically nested polygons, such as rectangles. Here, points of all the database trajectories are simply indexed by one single R-tree, while points from the same trajectory are further connected by a double linked list.

The trajectory route service 106 retrieves a nearest neighbor (λ−NN) of each geographical location of interest by using an incremental based k−nearest neighbor (k−NN) algorithm 602. This is assuming there is a set of geographical locations of interest of Q={q₁, q₂, . . . , q_m}, without specifying the order of travel for the multiple geographical locations of interest. The trajectory route service 106 retrieves the λ−NN of each geographical location of interest (λ>0) using the following:

$\begin{array}{c}\lambda -\mathrm{NN}\left(q_1\right)=\left\{p_1^1,p_1^2,\dots ,p_1^{\lambda }\right\}\\ \lambda -\mathrm{NN}\left(q_2\right)=\left\{p_2^1,p_2^2,\dots ,p_2^{\lambda }\right\}\\ \dots \\ \lambda -\mathrm{NN}\left(q_m\right)=\left\{p_m^1,p_m^2,\dots ,p_m^{\lambda }\right\}.\end{array}$

The trajectory route service 106 forms or creates a candidate set of trajectories from the GPS logs 604. A set of trajectories that have been scanned from the GPS logs contain at least one point in λ−NN(q_i) that is part of the candidate set C_i for identifying the “best” trajectory (k−BT) that connects each of the multiple geographical locations of interest. There may be several λ−NN points that belong to the same trajectory, thus a cardinality |C_i|≧λ may exist. The trajectory route service 106 merges the candidate sets that have been generated by all of the nearest neighbor searches λ−NN(q_i). As a result of the merging, there may be a possibility of very different trajectories as candidates for the best trajectory, based on the following:C=C₁∪C₂∪ . . . ∪C_m={R₁,R₂, . . . , R_f}where f is a number of trajectories. For each candidate trajectory R_x(∉[1, f]) that is within the candidate set C, the trajectory must contain at least one point whose distance to the corresponding geographical location of interest is determined. For example, if R_x∈C_i (C_iC), then the λ−NN of q_i must include at least one point on R_x, and the shortest distance from R_x to q_i is known. As a result, at least one matched pair of points between R_x and some q_i is identified. Thus, there may be a subset of trajectories from the candidate set that are matched to the at least the geographical location of interest specified by the user.

The trajectory route service 106 computes a lower bound LB of similarity function or proximal similarity for each candidate 606. The LB may be computed for each candidate R_x(∉[1, f]) by using the found matched pairs:LB(R_x)=Σ_{i∉[1,m]ΛRx∈Ci}(max_{j∉[1,λ]Λpij∈Rx}{e^{−Diste(qi,pij)}}).Here, {q_i|i∈[1, m]ΛR_x ∈ C_i} denotes a subset of geographical locations of interest that has already been matched with some point on R_x, and the p_i^j which achieves the maximum e^{−Diste(qi,pij)} with respect to q_i is the point on R_x that is closest to q_i. Thus, max_{j∉[1,λ]Λpij∈Rx}{e^{−Diste(qi,pij)}}=e^{−Diste(qi,Rx)}. The equation LB(R_x)=Σ_{i∉[1,m]ΛRx∈ Ci}(e^{−Diste(qi,Rx)}) is not greater than Σ_i=1^m e^{−Diste(qi,Rx)}, since it only takes the matched pairs that have been identified, into account. Thus, LB (R_x) may lowerbound the exact similarity Sim(Q, R_x) that is defined in Sim(Q, R)=Σ_i=1^m e^{−Diste(qi,R)}. However, if R_x ∉ C_i, then none of the trajectory points have been scanned by λ−NN(q_i).

The trajectories that are not contained in the candidate set C, are indicative that the trajectories have not been scanned by any of the nearest neighbor λ−NN searches, and any point on them may have a distance to qi no less than the distance of the λ^th NN of q_i (i.e., Dist_e(q_i, p_i^λ)). Therefore, the trajectory route service 106 computes an upper bound UB for of similarity function or proximal similarity for all of the non-scanned trajectories 608 (or trajectories that are not identified to be included in the candidate set). The equation to compute the upper bound UB is: UB_n=Σ_i=1^m e^{−Diste(qi,piλ}). Based on the results of the lower bound LB and the upper bound UB, the trajectory route service 106 applies a pruning mechanism to remove the unqualified candidates from the candidate set. The trajectory route service 106 identifies the trajectories that may be used as the best trajectory. The pruning mechanism is to avoid scanning the whole trajectory database which utilizes more time and cost. The search space for the trajectories is restricted by using the above method.

The trajectory route service 106 uses a theorem to determine if the number of best connected trajectory (k−BT) is included as part of the candidate set. The theorem is based on without specifying the order of travel to the geographical locations of interest. The trajectory route service 106 may receive a subset of a number of trajectories C′ from the candidate set C after searching the λ−NN of each geographical location of interest. The result found may be min_Rx∈C′{LB(R_x)}≧UB_n, then the number of best connected trajectories may be included in the candidate set C. The proof is for any R_x ∈C′,Sim(Q,R_x)≧LB(R_x) and for any R_y ∉ C (i.e. R_y∈C), UB_n≧Sim(Q, R_y). When min_Rx∈C′{LB(R_x}≧UB_n, the trajectory route service 106 may determine that:∀R_x∀R_y(R_x∈C′ΛR_y∉C)→(Sim(Q, R_x)≧Sim(Q,R_y)).Based on this, the connected trajectories result may not be from C but are from the candidate set C.

The trajectory route service 106 updates k maximal lower bounds, k−LB[ ] ⊂ LB[ ]. The trajectory route service 106 determines if the theorem is satisfied at 610. If the theorem is satisfied, then the k−BT that is included in the candidate set and the non-scanned trajectories beyond the candidate set may be safely filtered. Then the trajectory route service 106 proceeds to 612.

The trajectory route service 106 refines the candidates from the candidate set 612. Detailed discussion of the refining follows in FIG. 7.

Returning to 610, if the theorem is not satisfied, the process moves to the right side 614. If the best connected trajectory k−BT is not found in the candidate set, the trajectory route service 106 increases λ by a Δ 614 for the trajectory searches to locate or to ensure that the best connected trajectory is contained in the candidate set. If λ is set to be a very large value, the possibility is that the connected trajectories results will all be retrieved, but the search space may be huge, which may take a longer time period. However, a smaller λ may not be sufficient to ensure that the connected trajectories results are included in the candidate set, leading to a false dismissal. Rather, than choosing a fixed λ, the trajectory route service 106 applies an incremental number of nearest neighbor algorithm by increasing λ by a Δ for a next round of iterations. The process returns to 602 and starts another iteration. This k−NN algorithm provides an efficient retrieval of the candidate trajectories with a filtering and refinement mechanism.

The k−NN algorithm for computing, refining, and pruning steps of FIG. 6 is shown below:

k-NN Algorithm Identify Connected Trajectories
Input: k,Q
Output: k-BT
1.   Candidate Set C;
2.  Upperbound UBn;
3.  Lowerbounds [LB[ ], k − LB[ ]]
4.  Integer λ← k;
5.  While true do
6.   | For each qi ∈ Q from q1 toqm do
7.   | |  λ −NN(qi) ← KNN(qi,λ);
8.   | |  Ci ← trajectories scanned by λ −NN(qi);
9.   | C ← C1 ∪ C2 ...∪ Cm
10.  |  if |C| ≧ k then
11.  |  | compute LB [ ] for all trajectories in C;
12.  |  | k-LB[ ]←LB[ ].topK( );
13.  |  | if k-LB[ ].min≧UBn then
14.  |  | | k-BT← refine (C);
15.  |  | | return k-BT;
16.  | λ←λ + Δ;

FIG. 7 illustrates an illustrative process 612 of refining candidates from the candidate set. The trajectory route service 106 maintains a list of the best connected trajectories identified and the similarity values 702 from the process described in FIG. 6.

The trajectory route service 106 computes the UB for each candidate in the candidate set 704. The equation to compute the UB for each candidate is:

$\mathrm{UB}\left(R_x\right)=\sum _{i\in \left[1,m\bigwedge R_x\in C_i\right]}\underset{j\in \left[1,\lambda \right]\bigwedge p_i^j\in R_x}{\max }\left\{e^{-{\mathrm{Dist}}_e\left(q_i,p_i^j\right)}\right\}+\sum _{i\in \left[1,m\right]\bigwedge R_x\notin C_i}\left(e^{-{\mathrm{Dist}}_e}\left(q_i,p_i^{\lambda }\right)\right)$ where R_x ∉ C={C₁ ∪ C₂, . . . , ∪ C_m}. For a geographical location of interest within q_i|i ∈[1,m]R_x ∈ C_i, the closest point on R_x to is found by the λ−NN(q_i) search, and accumulate to UB(R_x), the contribution of the matched pair, q_i,closestPoint. Otherwise, for a q_i that the nearest neighbor search has not covered any point on R_x (i.e. R_x∉ C_i), the trajectory route service 106 considers that the current λ^th NN of q_i (i.e. p_i^λ) may be closer than the matched point, and accumulate the contribution of the q_i,p_i^λ pair to UB (R_x). Thus, the similarity or proximal similarity may be defined as:

$\mathrm{Sim}\left(Q,R_x\right)-\mathrm{UB}\left(R_x\right)=\sum _{i=1}^m{e}^{-{\mathrm{Dist}}_q\left(q_i,R_x\right)}-\sum _{i\in \left[1,m\right]\bigwedge R_x\in C_i}\left(e^{-{\mathrm{Dist}}_q\left(q_i,R_x\right)}\right)-\sum _{i\in \left[1,m\right]\bigwedge R_x\notin C_i}\left(e^{-{\mathrm{Dist}}_e\left(q_i,p_i^{\lambda }\right)}\right)=\sum _{i\in \left[1,m\right]\bigwedge R_x\notin C_i}\left(e^{-{\mathrm{Dist}}_q\left(q_i,R_x\right)}-e^{-{\mathrm{Dist}}_e\left(q_i,p_i^{\lambda }\right)}\right)\le 0.$ For any candidate R_x within C, the similarity function or the proximal similarity may be shown as Sim(Q, R_x)≦UB(R_x). The algorithm for refining the candidate set is shown below.

The trajectory route service 106 sorts the candidates from the candidate set in a descending order of UB 706.

The trajectory route service 106 determines whether the minimum similarity of the best connected trajectories is greater than or equal to the UB of the next trajectory candidate, R_x+1 708. If this occurs, the trajectory route service 106 identifies the trajectories as part of being included in the best connected trajectories 710. The trajectory route service 106 returns the connected trajectories as a final result.

The algorithm to compute refining candidates from the candidate set 612 is shown below:

Algorithm Refine Candidate Set C
1.  k-BT← SortedList(k);
2.  compute UB for each candidate in C;
3.  sort candidates in C by UB in descending order;
4.  for x=1 to |C| do
5.   | compute Sim (Q,Rx) by traversing Rx ;
6.   | if x ≦ k then k-BT.insert (Rx,Sim(Q,Rx));
7.   | else
8.   |  | if x = |C| or k − BT.min ≧ UB(Rx+1) then
9.   |  |  return k-BT;
10.  |  | if Sim(Q,Rx) > k − BT.min then
11.  |  |  | k-BT.removedLast( );
12.  |  |  | K-BT.insert (Rx,Sim(Q,Rx)).

Construct Trajectory Route and Examples of Trajectory Routes

FIGS. 8-10 illustrate an illustrative process 208 of constructing the trajectory route and illustrative trajectory routes. In FIG. 8, the process 208 constructs the trajectory or the travel route that allow the user 108 to travel in the order specified, if requested.

The trajectory route service 106 accesses the trajectory route service 802. A trajectory route map is constructed from global position system (GPS) logs associated with respective individual devices, each of the GPS logs include trajectories that connect a set of geographical locations previously visited by an individual of a respective individual device.

The trajectory route service 106 receives a request from the user 108 for a route to multiple geographical locations of interest 804. The user 108 may enter the request by selecting the multiple geographical locations of interest on the trajectory route map. In another implementation, the user may specify the multiple geographical locations of interest by entering input on the UI 114. As mentioned, the request may be described as a set of geographical locations of interest.

The trajectory route service 106 computes an initial path by identifying trajectories that are closest to each of the multiple geographical locations of interest 806. The trajectory route service 106 refines the initial path by finding the best trajectory 808 from the GPS logs that sequentially connects each of the multiple geographical locations of interest. The refining process was described in FIG. 7. The trajectory route service 106 presents the trajectory route with a travel sequence 810 based at least in part on connecting the multiple geographical locations of interest.

As discussed previously, the user 108 may specify a traveling order. When the order is specified, the trajectory route service 106 marks the travel route to allow the requesting user to view the route and to travel in the order specified 208.

FIG. 9 illustrates an illustrative trajectory route 900 of multiple geographical locations of interest without specifying any particular travel order by the user 108. For example, the user 108 inputs locations of for a “camp site” 902 that is located in Colorado near the vicinity of a town known as Buchanan, “North Sterling State Park” 904 to identify a park area for activities, such as hiking or boating, and “nearest highway” 906 to find a shortest distance route to travel from North Sterling State Park to a road. This travel route may be given by searching the trajectories from the GPS logs being closest in distance to each of the multiple geographical locations of interest 902, 904, and 906.

In an implementation, the trajectory route service 106 accesses a trajectory route model constructed from global positioning system (GPS) trajectories and geographical regions and receives user input to identify multiple geographical locations by the user clicking on a trajectory route map. The trajectory route service 106 computes the initial trajectory path based on a first geographical location to a second geographical location by using the trajectories that are closest in distance to the first and the second geographical locations. The trajectory route service 106 computes a secondary trajectory path based on the second geographical location to a third geographical location by using the trajectories that are closest in distance to the second and the third geographical locations, and refines the initial and the secondary trajectory paths by computing a trajectory route that sequentially connects the initial and the secondary trajectory paths.

The trajectory route service 106 adapts the k−NN algorithm to find the best trajectory with respect to the order of travel specified by the user 108. Using the candidate trajectory R_x ∈ C that is generated by the k−NN algorithm, some of the trajectory points are scanned by the λ−NN searches. For a set of scanned points on R_x by R′_x, the equation shows:R′_x={p_i|p_i∈R_xΛp_i∈S}

where S=λ−NN(q₁)∪λ−NN(q₂)∪ . . . ∪λ−NN(q_m). The R′_x, is a sub-trajectory that includes only a subset of points on R_x. The trajectory route service 106 allows R′_x, following the order of R_x. The equation for order specified similarity function is Sim_o(Q, R_x)≧Sim_o(Q, R′_x). The trajectory route service 106 uses another equation to calculate a new lower bound LB_o of similarity for ordered geographical locations by using a partially retrieved trajectory points of R_x. The equation for calculating the LB_o follows:LB_o(R_x)=Sim_o(Q,R′_x)=DP(Q,R′_x)where DP(Q, R′_x) is calculated using the algorithm shown below.

Algorithm for Ordered Travel: DP(Q,R)
1  Matrix M[i,j];
2  ∀∈ [1,m],M[i,0] ← 0
3  ∀∈ [1,l],M[o,j] ← 0
4  for i=1 to m do
5  | for j=1 to l do
6  | | if e−Diste(Head(Q),Head(R)) + M[i − 1,j] > M[i,j − 1]
| | then
| |  | // match qi with pj and repeat pj
7  | |  | M[i,j] ←
| |  | e−Diste(Head(Q),Head(R)) + M[i − 1,j]
8  | | else
| |  | // skip pj
9  | |  | M[i,j] ← M[i,j − 1]
10 Return M[m,l]

The trajectory route service 106 refines the process by calculating the UB_o (for ordered travel) for the candidate trajectories within the candidate set. The equation to use is:

${\mathrm{UB}}_o\left(R_x\right)={\mathrm{LB}}_o\left(R_x\right)+\sum _{i\in \left[1,m\right]\bigwedge R_x\notin C_i}\left\{e^{-{\mathrm{Dist}}_e\left(p_i^{\lambda },q_i\right)}\right\}$ where the k−NN algorithm may be adapted to find the best trajectory for the multiple geographical locations when the order of travel is specified by the user.

FIG. 10 illustrates an illustrative route 1000 provided by the trajectory route service 106. The user 108 specifies tourist attractions (e.g., landmark or famous name of attraction) by specifying an order of travel. The multiple locations are identified by landmarks, such as location 1, L1 is “The Capitol” 1002, location 2, L2 is “J. Edgar Hoover FBI Building” 1004, and location 3, L3 is “The National Mall” 1006. These are popular tourist attractions located in Washington, D.C. that a tourist, the user, or an agency could enter as input to receive the travel sequence in this order of locations 1, 2, and 3 (L1, L2, L3) based on the route along one way streets, on a bus route, or part of a travel agency's route.

In an implementation, the trajectory route service 106 receives user input specifying an order of travel for the geographical locations. The order is from first to third to second geographical locations. The trajectory route model provides an initial route based on a sequence of trajectories for the geographical locations. The trajectory route service calculates a new trajectory path based at least in part on using the points on the trajectories that are closest in distance to a first geographical location and a third geographical location. The trajectory route service calculates another new trajectory path based at least in part on the points on the trajectories that are closest in distance to the third and a second geographical locations, and refines the new trajectory and another new trajectory paths by computing the trajectory route based at least in part on the order of travel specified by the user that sequentially connects the first, the third, and the second geographical locations.

Exemplary Server Implementation

FIG. 11 is a block diagram showing an illustrative server usable with the environment of FIG. 1. The trajectory route server 112 may be configured as any suitable system capable of services, which includes, but is not limited to, implementing the trajectory route service 106 for online services, such as providing recommendations for the trajectory route, a trip planner, and the like. In one illustrative configuration, the server 112 comprises at least one processor 1100, a memory 1102, and a communication connection(s) 1104. The communication connection(s) 1104 may include access to a wide area network (WAN) module, a local area network module (e.g., WiFi), a personal area network module (e.g., Bluetooth), and/or any other suitable communication modules to allow the server 112 to communicate over the network(s) 104.

Turning to the contents of the memory 1102 in more detail, the memory 1102 may store an operating system 1106, a module for the trajectory route application 110, a trajectory route model module 1108, a connected trajectory module 1110, and an order travel sequence module 1112. Furthermore, there may be one or more applications 1114 for implementing all or a part of applications and/or services using the trajectory route service 106. The applications 1114 may be for implementing other programs, such as email, voicemail, and the like.

The trajectory route service 106 provides access to the trajectory route application 110. The functions described may be performed by the trajectory route service 106 and/or the trajectory route application 110. The trajectory route service 106 receives the user queries, sends the routes, builds the model, constructs the route, and interacts with the other modules to provide directions with sequence for travel.

The trajectory route application module 110 interacts with the trajectory route service 106. It provides the display of the application on the user interface, interacts with information from the trajectory maps, models, and other modules to provide recommendations for travel.

The trajectory route model module 1108 preprocesses the GPS data (or other location based logs) to identify points on the trajectory of the GPS logs. The process includes collecting or receiving GPS logs, parsing trajectories from the log data, and identifying trajectories that have a proximal similarity to the geographical locations.

The connected trajectory module 1110 determines the trajectories from the GPS logs that are similar to the geographical locations of interest, determines the best trajectory that connects each of the geographical locations, and provides the trajectory route. The connected trajectory module 1110 applies the algorithms described.

The order of travel sequence module 1112 correlates the order of travel specified by the user 108. The order of travel sequence module 1112 identifies the travel sequence by reordering the sequence of travel for the trajectory points based on using the algorithms described above.

The server 112 may include a trajectory route database 1116 to store the collection of GPS logs, trajectories, data for the trajectory route model, and the like.

The server 112 may also include additional removable storage 1118 and/or non-removable storage 1120. Any memory described herein may include volatile memory (such as RAM), nonvolatile memory, removable memory, and/or non-removable memory, implemented in any method or technology for storage of information, such as computer-readable storage media, computer-readable instructions, data structures, applications, program modules, emails, and/or other content. Also, any of the processors described herein may include onboard memory in addition to or instead of the memory shown in the figures. The memory may include storage media such as, but not limited to, random access memory (RAM), read only memory (ROM), flash memory, optical storage, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by the respective systems and devices.

The server as described above may be implemented in various types of systems or networks. For example, the server may be a part of, including but is not limited to, a client-server system, a peer-to-peer computer network, a distributed network, an enterprise architecture, a local area network, a wide area network, a virtual private network, a storage area network, and the like.

Various instructions, methods, techniques, applications, and modules described herein may be implemented as computer-executable instructions that are executable by one or more computers, servers, or telecommunication devices. Generally, program modules include routines, programs, objects, components, data structures, etc. for performing particular tasks or implementing particular abstract data types. These program modules and the like may be executed as native code or may be downloaded and executed, such as in a virtual machine or other just-in-time compilation execution environment. The functionality of the program modules may be combined or distributed as desired in various implementations. An implementation of these modules and techniques may be stored on or transmitted across some form of computer-readable media.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are disclosed as illustrative forms of implementing the claims.

Claims

1. A method implemented at least partially by a processor, the method comprising: receiving global position system (GPS) logs from user devices associated with GPS sensors;accessing, from a database storing the GPS logs, a first trajectory having a plurality of points, wherein an individual point in the plurality of points identifies a geographic location previously visited by a first user with a first user device of the user devices;generating a first user interface to provide functionality for a second user;causing the first user interface to be presented to the second user via a display of a second user device of the user devices associated with the second user, the first user interface: including a first map illustrating the geographic location, andconfigured to receive input from the second user;receiving a set of desired geographical locations from the second user via the first user interface, the first map illustrating at least one location of the set of desired geographical locations;receiving a request for a second trajectory associated with the set of desired geographical locations;determining that the first trajectory is a candidate trajectory based at least in part on: calculating, by the processor, a spatial distance between an individual point of the plurality of points and a desired geographical location of the set of desired geographical locations;calculating, by the processor and based at least in part on applying a similarity function to at least the spatial distance, a proximal similarity between the first trajectory and the set of desired geographical locations; anddetermining that the proximal similarity is less than a predetermined threshold;accessing additional trajectories to create a set of candidate trajectories for determining the second trajectory, wherein the set of candidate trajectories includes the first trajectory;removing unqualified candidate trajectories from the set of candidate trajectories based at least in part on proximal similarities determined between the set of desired geographical locations and individual unqualified candidate trajectories of the unqualified candidate trajectories, wherein resulting candidate trajectories comprise a refined set of candidate trajectories;identifying the second trajectory from the refined set of candidate trajectories, wherein the second trajectory is determined based at least in part on a latitude of the desired geographical location determined by a GPS sensor of the first user device, and a longitude of the desired geographical location determined by the GPS sensor;generating a second user interface to provide functionality for presenting the second trajectory to the second user; andcausing the second user interface to be presented to the second user via the display of the second user device, the second user interface including a second map illustrating: the set of desired geographical locations,the second trajectory connecting each location of the set of desired geographical locations, andan ordered travel sequence corresponding to the second trajectory.

2. The method of claim 1, wherein receiving the request for the second trajectory is based at least in part on the second user selecting each of the desired geographical locations of the set of desired geographical locations on the first map.

3. The method of claim 1, wherein the set of desired geographical locations is approximately ten or less geographical locations.

4. The method of claim 1, wherein removing the unqualified candidate trajectories comprises: computing a lower bound of proximal similarity for individual candidate trajectories in the set of the candidate trajectories;computing an upper bound of proximal similarity for candidate trajectories that are external to the set of the candidate trajectories; andremoving the unqualified candidate trajectories based at least in part on the lower bound of proximal similarity and the upper bound of proximal similarity.

5. The method of claim 1, wherein identifying the second trajectory is based at least in part on receiving another request specifying an order of travel for the set of desired geographical locations.

6. The method of claim 1, wherein identifying the second trajectory is based at least in part on historical data of a travel sequence of the second user.

7. The method of claim 1, wherein the removing unqualified candidate trajectories from the set of candidate trajectories is further based at least in part on an incremental based k-nearest neighbor analysis.

8. The method of claim 7, wherein removing the unqualified candidate trajectories from the set of candidate trajectories based at least in part on the incremental based k-nearest neighbor analysis comprises: indexing individual points in individual candidate trajectories of the set of candidate trajectories in a spatial index;determining that at least one individual point of the individual points associated with an individual candidate trajectory of the set of candidate trajectories in the spatial index is outside of a threshold distance from individual desired geographical locations of the set of desired geographical locations; andremoving the individual candidate trajectory of the set of candidate trajectories.

9. The method of claim 1, wherein identifying the ordered travel sequence is based at least in part on at least one of bus routes, traffic flow patterns on one way streets, or traffic flow.

10. The method of claim 1, wherein the second trajectory is determined based at least in part on a time associated with the first user device being located at the desired geographical location determined by the GPS sensor.

11. The method of claim 1, wherein the second trajectory is determined based at least in part on respective latitudes and longitudes of each geographical location in the set of desired geographical locations, the respective latitudes and longitudes being determined by at least one of the GPS sensors.

12. The method of claim 8, wherein the indexing the individual points in the individual candidate trajectories comprises indexing the individual points in a rectangle-tree (R-tree).

13. A system comprising: a processor;a memory coupled to the processor and storing: a trajectory route application module to receive input specifying a first set of geographical locations of interest to a first user; anda trajectory route model module to: receive location-based logs that are received from global position system (GPS) sensors associated with individual user devices;access individual location-based logs of the location-based logs from a database storing the location-based logs;generate a first user interface;cause the first user interface to be displayed via a display of an individual user device of the first user, the first user interface including a first map configured to receive input from the first user;receive the input specifying the first set of geographical locations from the first user via the first user interface, the first map illustrating at least one location of the first set of geographical locations;determine trajectories connecting a second set of geographical locations, wherein the second set of geographical locations were previously visited by a second user with an additional user device including a GPS sensor;determine, based at least in part on a similarity function, a proximal similarity between a first individual geographical location of the first set of geographical locations and a second individual geographical location from the location-based logs, wherein the location-based logs include a latitude of the second individual geographical location determined by the GPS sensor of the additional user device, and a longitude of the second individual geographical location determined by the GPS sensor;determine a plurality of initial paths, individual initial paths of the plurality of initial paths connecting the first set of geographical locations based at least in part on the proximal similarity;refine the plurality of initial paths based at least in part on the proximal similarity, the refining comprising: arranging coordinates associated with individual geographical locations of the second set of geographical locations into a spatial index;based at least in part on the spatial index, creating a candidate set of trajectories from the plurality of initial paths, the candidate set of trajectories including at least the first individual geographical location of the first set of geographical locations; andpruning unqualified candidate trajectories from the candidate set of trajectories to form a refined set of candidate trajectories based at least in part on determining that proximal similarities associated with the unqualified candidate trajectories are above a predetermined threshold;identify a particular trajectory from the refined set of candidate trajectories, wherein the particular trajectory is determined based at least in part on the latitude and the longitude;generate a second user interface; andcause the second user interface to be displayed via the display of the individual user device of the first user, the second user interface including a second map illustrating: the first set of geographical locations,the particular trajectory connecting each location of the first set of geographical locations, andan ordered travel sequence corresponding to the particular trajectory.

14. The system of claim 13, wherein the determining the individual initial paths comprises using a k-nearest neighbor (k-NN) algorithm to: identify a subset of trajectories from the location-based logs that are closest to a geographical location of the first set of geographical locations based at least in part on searching the location-based logs using a k-nearest neighbor (k-NN) search;calculate a lower bound of proximal similarity for a particular trajectory in the subset of trajectories that is closest in distance to the first set of geographical locations; andcalculate an upper bound of proximal similarity for a particular trajectory in the subset of trajectories that is not identified in the subset of trajectories.

15. The system of claim 13, wherein: the trajectory route application module is further configured to receive another input specifying an order of travel for the first set of geographical locations; andthe trajectory route application module is further configured to adjust the individual initial paths by connecting the first set of geographical locations based at least in part on the order of travel.

16. The system of claim 13, wherein: the trajectory route model module is further configured to refine the plurality of initial paths by finding a trajectory from the location-based logs that sequentially connects the geographical locations of the first set of geographical locations; andthe trajectory route model module is further configured to provide the ordered travel sequence by using the trajectories from the location-based logs based at least in part on a highest calculated proximal similarity.

17. One or more computer storage devices storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: generating a first user interface;causing the first user interface to be displayed via a display of an individual user device of a first user, the first user interface including a first map configured to receive input from the first user;receiving input from the first user, via the first user interface, specifying a first set of geographical locations of interest to a first user, the first map illustrating at least one location of the first set of geographical locations;accessing location-based logs from a database storing the location-based logs, wherein the location-based logs are received from individual user devices associated with global position system (GPS) sensors and include trajectories connecting a second set of geographical locations previously visited by a second user with an additional user device including a GPS sensor;computing a path connecting the first set of geographical locations based at least in part on calculating a proximal similarity between a first individual geographical location of the first set of geographical locations and a second individual geographical location of the second set of geographical locations, wherein the location-based logs include a latitude of the second individual geographical location determined the GPS sensor of the additional user device, and a longitude of the second individual geographical location determined by the GPS sensor, the computing comprising: indexing GPS location information associated with the second set of geographical locations into a spatial index;based at least in part on the indexing, creating a candidate set of trajectories from the location-based logs, the candidate set of trajectories from the location-based logs including at least the first individual geographical location of the first set of geographical locations; andpruning unqualified candidate trajectories from the candidate set of trajectories to form a refined set of candidate trajectories based at least in part on determining that proximal similarities associated with the unqualified candidate trajectories are above a predetermined threshold;identifying a particular trajectory from the refined set of candidate trajectories, wherein the particular trajectory is determined based at least in part on the latitude and the longitude;generating a second user interface; andcausing the second user interface to be displayed via the display of the individual user device of the first user, the second user interface including a second map illustrating: the first set of geographical locations,the particular trajectory connecting each location of the first set of geographical locations, andan ordered travel sequence corresponding to the particular trajectory.

18. The one or more computer storage devices of claim 17, wherein the computing the path comprises using a k-nearest neighbor (k-NN) algorithm to: identify a subset of trajectories from the candidate set that are closest to a geographical location of the first set of geographical locations based at least in part on searching the spatial index using a k-nearest neighbor (k-NN) search;calculate a lower bound of proximal similarity for a particular trajectory in the subset of trajectories that is closest in distance to the first set of geographical locations; andcalculate an upper bound of proximal similarity for a particular trajectory in the subset of trajectories that is not identified in the subset of trajectories.

19. The one or more computer storage devices of claim 17, wherein the operations further comprise: receiving another input specifying an order of travel for the first set of geographical locations; andadjusting the path by connecting the first set of geographical locations in the order of travel specified by the user.

20. The one or more computer storage devices of claim 17, wherein the operations further comprise: refining the path by determining a trajectory from the location-based logs that sequentially connects geographical locations included in the first set of geographical locations; andproviding a travel route by using the trajectory from the location-based logs based at least in part on a highest calculated proximal similarity.